A. Dotti1, J. Apostolakis, G. Folger, V. Grichine,V. Ivanchenko, M. Kosov, A. Ribon,

V. Uzhinskiy, D. H. Wrightfor the Geant4 Hadronic Working Group

14th International Conference on Calorimetry in High Energy Physics (CALOR2010) - 10-14 May 2010, Beijing

RECENT IMPROVEMENTS ON THE DESCRIPTION OF HADRONIC INTERACTIONS IN GEANT4

1: CERN, PH-SFT : andrea.dotti@cern.ch1

A. Dotti CALOR2010

Overview

Introduction

Status of the simulation of hadronic interactions for HEP experiments

Recent developments

FTF based physics lists improvements

CHIPS one model physics list

Results: comparison of different physics lists of calorimetric quantities

2

A. Dotti CALOR2010

Introduction

Hadronic physics in Geant4:cross sections and models for hadron-nucleus interaction up to TeV For neutrons from thermal energies to TeV

Models are tuned with thin target data (not calorimeters test-beam)Models are assembled in physics lists: stable configurations (few billions of events simulated)

Example: QGSP_BERT (used at LHC since 3-4 years)Experiments compare physics lists with test-beam dataWe use simplified calorimeters to study the impact of hadronic models on calorimeter observables

“Geant4 is a toolkit for the simulation of the passage of particles through matter. Its areas of application include high energy, nuclear and accelerator physics, as well as studies in medical and space science”NIM A506 (2003) 250-303, IEEE TNS 53 No.1 (2006) 270-278

http://www.cern.ch/geant4

3

A. Dotti CALOR2010

Models Summary

Bertini cascade: low energy intra-nuclear cascade (best agreement with data up to Ekin ≈ 5 GeV) Nucl. Instr. Meth, 66, 1968, 29 ; Physical Review Letters 17, (1966), 478-481

Quark-Gluon-String, “QGS”: p,n,k,π of high energy (agreement with data from Ekin≈10-15 GeV) See Sec. IV, Chap. 22 of Geant4 Physics Reference Manual and bibliography within

Parameterized models (derived from GHEISHA as first hadronic model mid-90): all E and particles. Goal: replace with more accurate models. Still used in most physics lists for hyperons and antibaryons

Chiral Invariant Phase Space Decay,“CHIPS” (new developments): all E and particles. Eur. Phys. J. A 8, 217-222 (2000) ; Eur. Phys. J. A 9, (2001) ; Eur. Phys. J. A 9, (2001)

Fritjof, “FTF” (new developments): p,n,k,π of high energy (valid from Ekin≈4-5 GeV) Nucl. Phys. 281 289 (1987)

Rece

nt Im

prov

emen

ts

4

A. Dotti

Thin Target Tuning: Example

CALOR2010 5

Tuning is done at model level:Thin target dataSeveral tests are run routinely to follow evolution of model code

BERT and FTF predictions for HARP-CDP data: double differential cross sections in pA interactionsBERT and FTF describe data reasonably well

FTF and BERT can be used

together in the 4-5 GeV region

A. Dotti CALOR2010

Physics Lists In This TalkA Physics List is a set of consistent physics models for each particle in application

LHC tested several options: most challenging requirements on hadronic interactions come from ATLAS and CMS calorimeters

After detailed validation with test-beam: QGSP_BERT (2007)

For a given physics list when a hadronic interaction occurs a model, depending on primary type and energy, is sampled

http://geant4.cern.ch/support/proc_mod_catalog/physics_lists/physicsLists.shtml

6

QGSP_BERT

FTFP_BERT FTFPBERT4GeV 5GeV

BERT QGSPLEP9.5GeV 9.9GeV 12GeV 25GeV

p/nK/π

p/nK/π

CHIPS CHIPS

LEP

Models Physics Lists

LHEP

Theo

ry d

riven

allproj.

allproj.

CHIPS components used by other PL for capture of negative hadrons at rest

and γ-Nuclear and Lepto-Nuclear interactions

HEP

A. Dotti CALOR2010

Status Of Simulation

Need for precise simulation of observables for LHC (CERN-LCGAPP-2004-02):

Response (e/pi), resolution, shower shapesResults from ATLAS & CMS test beam:

best description obtained with QGSP_BERT physics list G4 9.3 under validation by experiments

Response: good agreement, within 3%Resolution: simulation is a bit too narrow, within 10%Showers still a bit shorter and narrower than data:

pions within 10% up to 10λprotons within 30% up to 10λ

See: JoP Conf. Series 160 (2009) 012073 ; CALOR08 Contribution by G. Folger

7

A. Dotti CALOR2010

Transition Between Models

CMS, and then ATLAS, observed unphysical steps in response as a function of beam energy

Confirmed with simplified setups

We have investigated in detail the source of this:

Related to transitions between models

Use of parametrized models (LEP) in medium (10-25GeV) energy range

This has been one of the main area of activity in Geant4 hadronic

Details presented at IEEE NSS/MIC 2009 - 25-31 October 2009 - A. Ribon’s contribution. Proceedings under publication

8

Simplified Cu/LAr calorimeter

9.9 GeV 25 GeV

A. Dotti CALOR2010

FTF Physics Lists Improvement Based on Fritiof model: string model with LUND fragmentation

Why FTF? Geant4 QGS is valid from Ekin ≈ 10-15 GeV, FTF is promising alternative: valid from Ekin≈4-5 GeV

Hadron-hadron interactions are modeled as binary reactions:

a’ and b’ are excited states of the initial hadrons a and b

FTF model in Geant4: simulation of single diffraction, simulation of binary reactions, Reggeon cascading

Recent improvements (quark exchange introduction, Reggeon cascading) allow for a smooth coupling with Bertini cascade models at 4-5 GeV: removing discontinuities

FTF is also implemented in HIJING, UrQMD, ART, HSD codes

Nucl. Phys. B 281 (1987) 289Comp. Phys. Comm. 43 (1987) 387

a + b→ a� + b� ; ma� > ma, mb� > mb

Original FTF Quark-exchange (new)

9

A. Dotti CALOR2010

New CHIPS Physics ListCHIPS model of inelastic nuclear interactions:

High energy: 1-D Parton Multi String (PMS), soft part of it absorbed by target nucleus creating quasmons

Low energy: 3-D decay of a quasmon (parton plasma) with final CHIPS evaporation

CHIPS physics list (released as experimental PL in Dec. 2009) very smooth: melds the HE and LE approaches

CHIPS can be used for all particles (including kaons, anti-baryons, hyperons)

EPJ A 8 (2000), 217EPJ A 9 (2000), 411EPJ A 9 (2000), 421

CHIPS provides also:

Revised cross sections for all hadrons

At rest nuclear capture processes for negative hadrons

Neutrino-nuclear, electron-, muon-, tau-nuclear and photo-nuclear reactions

Elastic scattering for all hadrons

Quasi-elastic scattering

10

A. Dotti CALOR2010

Simplified Fe/Sci Calorimeter

FTFP_BERT and CHIPS: smooth response. FTFP_BERT agrees with QGSP_BERT, where this one agrees with data

QGSP_BERT stable since G4 8.3 (May 2007)

FTFP_BERT smooth response (improved in G4 9.3)

CHIPS (new in G4 9.3) higher response

What about other observables (resolution, shower shapes)?

CHIPS still under tuning

π- beam geant4 9.3.ref02 (development version)

11

A. Dotti CALOR2010

Standard Deviation

Resolution (σ(Evis)/<Evis>) is not a good observable: <Evis> has steps, prefer to show σ(Evis)/Ebeam

CHIPS smaller width

QGSP_BERT: step at 10 GeV

12

π- beam geant4 9.3.ref02

FTFP_BERT and CHIPS: smooth.

A. Dotti CALOR2010

Bertini model (low energy) increases dimensions of showers

Small steps at 10 GeV (QGSP_BERT) and at 5 GeV (FTFP_BERT) are visible

CHIPS predicts longer showers at high energy

Feedback from experiments: agreement (QGS_BERT) with test-beam data has improved in the last years, models predicting longer and wider showers should be preferred (G4 still 10-30% shorter)

<r 2>

<λ2>Longitudinal Shower Shapes

Shower shapes: weighted average of read-out cells

position with respect to shower axis and shower center

Transitions< λ2 >=

�cell Ecellλ2

cell�cell Ecell

< r2 >=�

cell Ecellr2cell�

cell Ecell

13

π- beam geant4 9.3.ref02

A. Dotti CALOR2010

Lateral Shower Shape

Compared with data (E>20GeV): FTFP_BERT and CHIPS better agreement with data

FTF and CHIPS predicts wider showers at high energy and more compact at low energy

Cascading is a fundamental ingredient to increase shower size and thus improve agreement with test-beam data

14

Transitionπ- beam geant4 9.3.ref02

QGSP_BERT is smooth

FTFP_BERT has a step at 4-5 GeV (transition between BERT and FTFP): can be improved increasing transition region (under investigation)

A. Dotti CALOR2010

Conclusions

Thin target experiments are the primary source of data for tuning models

Agreement with test-beam data for response and resolution is O(few %)

Agreement for shower shapes have much improved but still a bit shorter and narrower with respect to test-beam data (worst case: shower length for protons -30% at 10λ),

Recent major improvements:

Fritiof model (available in FTFP_BERT physics list)

Extension of CHIPS components (available in CHIPS physics list)

Main concern is discontinuities in response

Studied in detail during 2009, origin tracked down to use of parametrized model (LEP) in intermediate region: now providing option with reduced or no dependence on this (FTFP_BERT, CHIPS)

FTFP_BERT is smooth and agrees with QGSP_BERT for Ekin<9 GeV and Ekin>25 GeV

15

A. Dotti CALOR2010

Challenges And Future Work

At the moment two physics lists show the most promising results to solve these issues:

FTFP_BERT physics list (Bert < 5 GeV ; FTF > 4 GeV). Transition effect to be corrected in shower dimension. Very similar results to QGSP_BERT at high and low energies

CHIPS physics list: very smooth. Response is too high, results should be considered preliminary: still “experimental” physics list, ongoing validation and tuning with thin target data

Expect further improvements in next months thanks to new data:

First comparison with LHC collisions

CALICE test-beams

Other challenges: improve simulation of hadronic interactions for kaons, anti-p, hyperons

Other possibilities being explored: example QGSP_FTFP_BERT (use of FTF instead of LEP)

16

A. Dotti CALOR2010

Thank you!

17

A. Dotti

BACKUP SLIDES

18

A. Dotti CALOR2010

Geant4 Hadronic Working Group

Dennis Wright (SLAC) - Working group coordinator

Gunter Folger (CERN) - Working group deputy coordinator

Makoto Asai (SLAC)

Sunanda Banerjee (Fermilab)

Alain Boudard (CEA)

G. A. Pablo Cirrone (INFN LNS)

Andrea Dotti (CERN)

Daniel Elvira (Fermilab)

Vladimir Grichine (CERN, Lebedev Physical Institute, Moscow)

Alexander Howard (CERN, ETH Zurich)

Vladimir Ivanchenko (CERN, EMSU Lomonosov Moscow State University)

Anton Ivanchtenko (IN2P3)

Fred Jones (TRIUMF)

Pekka Kaitaniemi (CEA, Helsinki Institute of Physics(HIP))

Michael Kelsey (SLAC)

Tatsumi Koi (SLAC)

Mikhail Kosov (CERN)

Fan Lei (QinetiQ)

José Manuel Quesada Molina (Universidad de Sevilla)

Alberto Ribon (CERN)

Francesco Romano (INFN LNS)

Pete Truscott (QinetiQ)

Vladimir Uzhinskiy (JINR Dubna)

Julia Yarba (Fermilab)

geant4-hadronics@cern.ch19

A. Dotti CALOR2010

Hadronic Inelastic Interactions Models

1 MeV 10 MeV 100 MeV 1 GeV 10 GeV 100 GeV 1 TeV

FTF String

QGS String

HEP

LEP

Bertini Cascade

Binary Cascade

Fission

Rad. Decay

γ de-excitation

Multifragment

Fermi breakup

Evaporation

Pre-Compund

Photo-nuclear, lepto-nuclear (CHIPS)

CHIPS 3-D quasmon decay CHIPS string

At restabsorptionanti-p, K-, μ-, π-

20

A. Dotti CALOR2010

Simplified CalorimeterSimulation of a cylindrical calorimeter 10 λI long

All LHC calorimeters technologies are implemented: Pb/Sci, Fe/Sci, Cu/LAr, PbWO4

Energy in active material is collected

x

y

z

proton shower with Ekin=50 GeVy

21

A. Dotti CALOR2010

Corrections And SystematicsThere are two effects introducing a systematic error in the simulation of energy response: late energy deposits (slow neutrons) and finite dimension of simplified calorimeterA time cut at 50 ns (typical read-out timing of scintillator calorimeters) have been introduced: deposited energy depends softly from this cut (3% varying from 20 to 200 ns)

1 5 10 50 100 500Ebeam�GeV�

0.02

0.03

0.04

0.05

0.06

0.07

0.08

Eleakage

Ebeam

Neutrino and muon included! Effect on visible energyis reduced to less than half

Albedo

Leakage

Most important correction is leakage (front and longitudinal): O(1%), error on leakage is a fraction of stat errorAt low energy correction for (front)leakage becomes important

22

A. Dotti

Resolution Vs WidthResponse and width extracted from iterative gaussian fit around response peak of visible energy

5 10 15 20Ebeam�GeV�

0.1

0.2

0.3

0.4

0.5Σ�E

QGSP_BERT Fit�

a

Ebeam⊕ b

NIMA606: a=(52.9±0.9)% ; b=(5.7±0.2)%

0 5 10 15 20Ebeam�GeV�

0.95

1.00

1.05

1.10

Σ �EFit

Modulation of residuals following transition regions: is this a problem in σ or

in <E>? Need to look at both independently

CALOR2010 23

A. Dotti CALOR2010

Shower Shape1. Define “mesh” of voxels: pseudo-cells

2. Accumulate quantity for each voxel (energy deposit, energy density)

3. If an observable O can be defined for each voxel (e.g. energy, density, position) the n-th moment in O can be calculated:

4. Moments are often calculated w.r.t. shower center (<x>,<y>,<z>) and shower axis (principal component analysis)

< On

>=�

v Ev · Onv�

v Ev, v ∈ voxels

-100 -50 0 50 100-80

-60

-40

-20

0

20

40

60

80

hEdep_xzEntries 10453Mean x -56.58Mean y -1.221RMS x 18.65RMS y 8.072

0

200

400

600

800

1000

1200

1400

1600

hEdep_xzEntries 10453Mean x -56.58Mean y -1.221RMS x 18.65RMS y 8.072

Energy Deposit_xz

-100 -50 0 50 100-80

-60

-40

-20

0

20

40

60

80

hEdep_yzEntries 10453Mean x -56.58Mean y -1.723RMS x 18.65RMS y 9.246

0

200

400

600

800

1000

1200

1400

1600

1800

2000

hEdep_yzEntries 10453Mean x -56.58Mean y -1.723RMS x 18.65RMS y 9.246

Energy Deposit_yz

-80 -60 -40 -20 0 20 40 60 80-80

-60

-40

-20

0

20

40

60

80

hEdep_yxEntries 10453Mean x -1.221Mean y -1.723RMS x 8.072RMS y 9.246

0

500

1000

1500

2000

2500

3000

3500

hEdep_yxEntries 10453Mean x -1.221Mean y -1.723RMS x 8.072RMS y 9.246

Energy Deposit_yx

z

x

y

y

x

z

beam direction

Original Idea from:T. Barillari et al. ; Local Hadron

Calibration; 2008 (CERN); ATL-LARG-PUB-2009-001

24

A. Dotti CALOR2010

Shower ShapeShower moments can be used to “summarize” the shower shape in few numbers:

λcenter : depth of the shower “maximum” w.r.t. calorimeter front-face

λ2 : shower dimension along shower axis

r ,r2: shower dimension in plane ortogonal to shower axis

Long. shapes depend weakly from mesh size (<10%)

Lateral shower shape depends weakly (<10%) from mesh size only for 1cm<size<10 cm. Prefer r2 over r: less dependency on mesh size

�

�� � � ��

�

�

� � �

0 1 2 3Log10�V� �cm3�

0.8

0.9

1.1

1.2Ratio

λcenter/λcenter(5cm)

1 cm

vox

el

20 cm

vox

el

λ2/λ2(5cm)

�� �

�

�

�

�� �

�

0 1 2 3Log10�V� �cm3�

0.8

1.2

1.4

1.6

Ratio r/r(5cm)r2/r2(5cm)

Systematic error O(5%) on shower shapes due to choice of

a particular mesh size

25

A. Dotti Meeting

Shower Shape

Shower moments distributions present a long tail...

Whisker Plot

...however they are regular w.r.t. beam energy...

Stat. Errors x3

...stat. errors are small: plot mean and error on the mean Vs beam

energy

However they are very regular w.r.t. beam energy

QGSP_BERT

QGSP_BERT

26

A. Dotti CALOR2010

Additional Physics Lists

27

A. Dotti CALOR2010

Additional Physics Lists

28

A. Dotti CALOR2010

Additional Physics Lists

29

A. Dotti CALOR2010

Additional Physics Lists

30

A. Dotti CALOR2010

Additional Physics Lists

31